E3S Web of Conferences 124, 01039 (2019) https://doi.org/10.

1051/e3sconf/201912401039
SES-2019

Improvement of the technical excellence of multiparameter relay

protection by combining the signals of the measuring fault
detectors using artificial intelligence methods
A. L. Kulikov1, A. A. Loskutov1,*, and M. Mitrovic1
1 Power Supply and Power Electronics Nizhny Novgorod State Technical University n.a. R.E. Alekseev, Nizhny Novgorod, Russia

Abstract. The features of the dielectric rod defects and the rod-electrod defects of polymer insulators were
investigated by using the developed complex method of contactless remote diagnostics of high-voltage
insulators health, which includes the initial detection of local areas with elevated electric field gradients as
well as measurement of a set of partial discharges characteristics. The radiation of extra-large partial
discharges was detected and partial discharges features were investigated. According to the measurement
results, a diagram was constructed for the electro-physical processes that accompany the emission of extra-
large partial discharges; and the reasons for their occurrence were established.

1 Introduction by automated synthesis based on the theory of automata

were considered.
The increase in electricity demand, the complication of It is known to use methods of processing redundant
power systems and the use of modern technologies lead information for RP and, in particular, voting schemes for
to the need to improve relay protection (RP). The use of increasing reliability [10-11]. Technical solutions of RP
microprocessor technology in RP devices provides the using artificial intelligence and machine learning, as well
flexibility and versatility of their work. One of the as algorithms for recognition of electrical network
modern directions in the evolution of RP is the modes based on artificial neural networks (ANN) and
development of multiparameter (multidimensional) decision tree (DT), were discussed in [12-18]. However,
algorithms [1-6]. Their basis is the use of multiple artificial intelligence methods were not used to organize
parameters of measured currents and voltages in the the logical part of multiparameter RP.
recognition of the operation modes of the electrical
network. The redundancy of the information base of
such RP provides the increase in the technical excellence 3 Purpose of the research
of devices, including their sensitivity, reliability and
The purpose of the article is to study and develop
speed.
algorithms for the implementation of multiparameter RP
using artificial intelligence methods, as well as
2 Relevance algorithms that provide high levels of recognizability of
emergency modes and reliability of operation by
Increasing the sensitivity of relay protection is one of the organizing the logical part of the device. In particular,
actual tasks. This is especially important in recognizing the implementation of this purpose implies the precise
emergency modes in modern power supply systems, determination of the triggering parameters (settings
including the presence of distributed generation sources. values) of individual measuring faults detectors (FD), as
For example, the cause of low sensitivity (not well as the effective combination of signals in the logical
recognizable damage) may be the presence of contact part of the multiparameter RP. This takes into account
resistance in the place of short circuit currents (SC). the minimization of instrumental and calculative costs.
One of the effective methods of increasing the To form a multiparameter RP, a statistical approach is
sensitivity is the use of multidimensional used, as well as pattern recognition procedures using
(multiparameter) RP based on the information approach ANN and DT methods.
[1-4]. There are developments of multiparameter
protection, for example, for solving problems of long-
distance backup with SC behind a transformer [5], as 4 Statement of the task
well as for determining the response parameters and
The use of the information approach in RP involves the
combining the signals of individual starting organs [6-8].
multiple implementation of simulation of the electrical
In [9], the options for organizing the logical part of RP
network and the protected object in normal and

*
Corresponding author: [email protected]

© The Authors, published by EDP Sciences. This is an open access article distributed under the terms of the Creative Commons Attribution License 4.0
(http://creativecommons.org/licenses/by/4.0/).
E3S Web of Conferences 124, 01039 (2019) https://doi.org/10.1051/e3sconf/201912401039
SES-2019

emergency modes. The development of RP algorithms is indices "1" and "0" (table 1) indicate the parameters of
based on model experiments [1–9]. In the course of the positive and zero sequences.
research, the simulation model of the 220 kV electrical Table 1 includes the following coefficients that
network was used (Figure 1). It is implemented in the specify the range of variation of the voltage of power
Matlab/Simulink software package (Figure 2). The systems and the angle of the phase shift between them:
model consists of three sections of a 220 kV overhead
transmission line, two power sources and a short-circuit k1 = Es1 / U nom , k2 = Es 2 / U nom ,
connector through contact resistance in the reserve zone  = arg( Es1 / Es 2) (1)
of protection RP1. The accumulation of measured
current and voltage data was performed at the To obtain the parameters of the statistical
installation place of the RP1 at the beginning of the first distributions of normal and emergency modes,
section of the power transmission line (Figure 1). These simulation modeling implied 10 000 iterations, including
data are necessary for the implementation and research single-phase short-circuits along the third section of
of the developed RP. power line, which represents the backup zone of the
observed protection RP1.
Multi-parameter relay protection was organized with
the use of five separate one-dimensional measuring fault
detectors: FD1 - by current module; FD2 - by voltage
module; FD3 - by voltage phase; FD4 - by active power;
FD5 - by reactive power. The obtained statistics data in
the form of histograms for normal and emergency modes
of each FD is approximated by a normal (Gaussian)
distribution. On the graphs of the densities of the normal
Fig. 1. The equivalent circuit of the observed section of the distribution, the horizontal axes represent the
electrical network 220 kV. measurements of the observed parameters of the
measuring FD in normal and emergency modes along
power line 3 (Figure 3). The possibilities for recognizing
the normal and emergency modes of the electrical
network by a RP device are determined by the size of the
area of intersection of the statistical distribution curves.
Perfect, 100% recognizability, is carried out in the
absence of the intersection of these curves.

5 Proposed method
Fig. 2. Simulation model of the network segment in To implement the methods discussed below, it is
Matlab/Simulink. proposed to define the settings values by solving the
statistical two-hypothetical task. It implies the
The parameters of the elements of the simulation implementation of the hypothesis testing procedure
model are shown in table 1. It must be emphasized, that regarding the electrical network mode by each FD at the
during the operation of the network segment (Figure 1), analyzed time point t: H0 - if the measurement belongs
the parameters of the elements may deviate from the to the range of normal modes (output signal - "0"); H1 -
specified values in the design. According to this, in order if the measurement belongs to the area of emergency
to obtain the required statistics on the network operation modes (output signal - "1").
modes, the model parameters were divided into fixed
and interval, that are varied in the specified ranges. The

Table 1. Parameters of model elements.

Fixed Interval
Parameter Value Parameter Value Parameter Value
s1
Unom (kV) 220 X1 (Ohm/km) 25.09 (±1) lline 0.0001…0.9999
s1
L1 (km) 100 X0 (Ohm/km) 1.31 (±0.1) Rf (Ohm) 0.1…40
L2 (km) 100 X1s2 (Ohm/km) 9.75 (±1)  (˚) -30˚… 30˚
L3 (km) 100 X0s2 (Ohm/km) 0.87 (±0.1) K1 0.95…1.05
b11,2,3 (S/ km) j0.6510-6 R11,2,3(Ohm/km) 0.13 (±0.01) K2 0.95…1.05
b01,2,3 (S/ km) J0.6510-6 R01,2,3(Ohm/km) 0.27 (±0.01) P (MW) 300 (±10)
R1s1,s2 (Ohm/km) 0 X11,2,3 (Ohm/km) 0.332 (±0.01) Q (MVar) 150 (±10)
R0s1,s2 (Ohm/km) 0 X01,2,3 (Ohm/km) 0.332 (±0.01)

2
E3S Web of Conferences 124, 01039 (2019) https://doi.org/10.1051/e3sconf/201912401039
SES-2019

Fig. 3. Determination of setting values by detuning from normal modes: (a) FD1 - by current module, (b) FD2 - by voltage
module, (c) FD3 - by voltage phase, (d) FD4 - by active power, e) FD5 - by reactive power.

same curve belonging to hypothesis H0. Based on the

above, the curve of normal modes completely belongs to
5.1 Determination of the parameters of the
the hypothesis H0 (p0=0, q0=1).
relay protection, based on the areas of normal
and emergency modes
5.2 Combining of logic signals to RP
The main requirement in determining the settings values protection modes using an artificial neural
of PR is to prevent protection operation in all possible network
normal modes. The determination of the settings values
of each FD can be performed, for example, by the Baes The integration of logical signals at the outputs of
criterion for minimizing the average risk of decision measuring FD of RP is implemented using ANN and
making [7-8]. In this case, the setting value η must be set based on the Neural Net Pattern Recognition (nprtool)
to a value that corresponds to a full detuning from all tool in the Matlab software package. For the training of
normal modes. In them, the RP should not operate the neural network, 14 000 combinations of normal and
(Figure 3). emergency modes (70% of the total sample loaded for
experiments with ANN) were used. The combinations
Table 2. Probability of recognition of emergency modes of were randomly selected. Also, 3000 combinations (15%)
individual measuring fault detectors.
were used for validation and 3000 (15%) for testing. The
Measuring fault p1 q1 training was conducted on a simple neural network with
Parameter one hidden layer of ten sigmoidal neurons, an input layer
detectors (%) (%)
FD1 - by current module Im 73.78 26.22 of five neurons and an output layer of two "softmax"
FD2 - by voltage module Um 43.16 56.84 neurons (Figure 4). Figure 5 illustrates the structural
FD3 - by voltage phase φu 6.28 93.72 diagram for combining the logical signals of individual
FD4 - by active power P 12.14 87.86 measuring FD using ANN.
FD5 - by reactive power Q 56.07 43.93

To determine the probability of the zones of

triggering and not triggering belonging to the hypotheses
H0 (normal mode) and H1 (emergency mode), features of
the normal distribution of a continuous quantity were
used [19]. Table 2 presents the probability of operation
(p1) and non-operation (q1=1-p1) of each FD for the Fig. 4. Model of used ANN.
curve belonging to the emergency mode. The probability
of triggering p1 corresponds to the area under the
emergency mode curve belonging to hypothesis H1, and
the probability of non-triggering q1 - the area under the

3
E3S Web of Conferences 124, 01039 (2019) https://doi.org/10.1051/e3sconf/201912401039
SES-2019

observations of emergency and 10 000 observations of

normal modes were made. The structure of the obtained
decision tree and the signal combining scheme in the
logical part are shown in figures 7 and 8.

Fig. 5. The scheme of combining signals using ANN.

Training of the neural network model was achieved

in 25 iterations with a duration of about 1 ms. The final
confusion matrix (figure 6) displays the ratio of
recognizable and unrecognizable RP modes, as well as
the number of used experimental measurements of Fig. 7. Structure of the observed DT.
normal (10 000 = 50%) and emergency (10 000 = 50%)
modes.

Fig. 8. Scheme of combining signals by the DT method.

From the analysis of the decision tree structure

Fig. 6. ANN confusion matrix.
(figure 7), it was determined that when combining
The element of the matrix (1,1) represents the normal logical signals, only four measuring FD are sufficient.
modes recognized by ANN, and (2.2) - the emergency FD4 - by active power does not affect the increase in
modes. The element of the matrix (1,2) characterizes the recognizability of modes in the relay protection device.
modes that are considered emergency. But during This means that FD4 is also redundant in the formation
training, they turned out to be normal (incorrect of multiparameter relay protection in the RP1 location.
recognition of emergency modes by RP). The field (2.1) Monitoring active power is optional. Thus, protection is
corresponds to modes that were considered normal. But simplified and its reliability is additionally increased,
during training, they turned out to be emergency while maintaining high recognizability of emergency
(incorrect recognition of normal modes by relay modes.
protection). From the analysis of figure 6 it can be seen
that the complete detuning from the normal modes was
performed correctly (the field (2.1) is 0). When
combining signals in the logical part of multidimensional
RP, the recognizability of emergency modes has
increased and is 84.34% (the value of the matrix element
(2.2) is equal to 8434).

5.3 Combining logic signals from the outputs of

measuring FD using the decision tree method
The Matlab Classification learner application (fitctree)
was used to combine the signals from the measuring FD
and decide whether the relay protection was triggered or
non-triggered. The classification of signals was carried
out with 20 000 observations that make up the database Fig. 9. DT confusion matrix.
of parameters of the electrical network modes at the Calculations of the confusion matrix of the DT
installation site of the RP1 protection. 10 000 (figure 9) show that it corresponds to the ANN confusion

4
E3S Web of Conferences 124, 01039 (2019) https://doi.org/10.1051/e3sconf/201912401039
SES-2019

matrix. Diagonal fields (fault, fault) and (normal, n

normal) represent the correct recognition of emergency q(t ) =  qi (t )

and normal modes. The field (normal, fault) represents i =1 (3)
the number of modes in which the RP malfunctions. It
turned out to be zero. The number of unrecognized The statement of numerical values in expression (3)
emergency modes was 15.06% (fault, normal). Thus, leads to the probability of failure of the integrated relay
when combining logic signals from measuring FD using protection equal to q(t)=3.6∙10-5. On this basis, the
the DT method, protection will work correctly, and the probability of failure-free operation is p(t)=1-q(t)=1-
recognizability of emergency modes increases to 3.6∙10-5=0.99996≈100%.
84.34%. Thus, combining the methods of organizing the
logical part of relay protection based on the "1 of 2"
principle leads to a significant increase in the reliability
5.4 Combined method of relay protection.
To improve the reliability of multidimensional RP, it is
proposed to organize the logical part of a protection 6 Conclusions
device with a combination of ANN and DT methods
based on the "1 of 2" vote. To assess the reliability of 1. There have been analyzed the applicability of the
such a logical part, a model of failures of unrecoverable methods of combining the signals of the logical part of
systems was used. The combination of methods is multiparameter relay protection using artificial
carried out by parallel connection of their blocks intelligence: ANN, DT and voting schemes according to
(figure 10). the "1 of 2" principle. For the implementation of
multiparameter relay protection, there were used the
results of simulation of the electrical network and
statistical techniques for processing the results of model
experiments.
2. There was investigated a multiparameter relay
protection with five individual one-dimensional
measuring faults detectors. Measuring fault detector by
current module (FD1) has the highest recognizability of
emergency modes. The probability of recognition is
73.78%. The combination of logical signals using the
three proposed methods ensured an increase in the
probability of recognizing an emergency mode to
84.34%.
Fig. 10. The scheme of the logical part with the combined 3. The choice of the method for combining the output
method.
signals of the measuring faults detectors in the logical
Based on the fact that ANN and DT are considered as part of the multiparameter relay protection is dependent
reliable methods, we suppose that the probabilities of on the characteristics of the electrical network modes
reliable operation of the ANN and DT blocks (Figure 10) and the requirements for the organization of the
are the same. They correspond to the probabilities of protection of the electric power facility.
microprocessor logic solvers (logic solver) used on
The presented research results were obtained with the
objects of high security, for example, at nuclear power
support of a grant from the President of the Russian Federation
plants. for state support of young Russian scientists (MK-
It should be noted that the probability of reliable 3210.2019.8). Agreement No. 075-15-2019-337 of 11.06.2019.
operation of individual logical solvers is 99.4% during
the operation period [20].
The failure model of unrecoverable systems implies References
that: pi(t) is the probability of failure of the i-th system to
work; qi(t) is the probability of failure of the i-th system. [1] Y.Y. Lyamets, G.S. Nudel'man, D.V.
Assuming that the methods (systems) have the same Zinov'yev, D.V. Kerzhayev, Y.V. Romanov,
probability of failure-free operation during the Mnogomernaya releynaya zashchita. CH.1.
exploitation of RP, p1(t)=p2(t)=0.994. The probability of Teoreticheskiye predposylki Elektrichestvo 10, 17-25 (in
failure is defined as: Russian) (2009)
[2] Y.Y. Lyamets, G.S. Nudel'man, D.V.
qi(t)=1- pi(t), (2) Zinov'yev, D.V. Kerzhayev, Y.V. Romanov,
Mnogomernaya releynaya zashchita. CH.3.
As a result, it turns out that q1(t)=q2(t)=1- Ekvivalentirovaniye modeli Elektrichestvo 1, 9-15 (in
0.994=0.006. With parallel connection of logical blocks Russian) (2009)
of the ANN and DT the probability of failure of the [3] Y.Y. Lyamets, G.S. Nudel'man, I.S.
combined principle "1 of 2" of the system is Podshivalina, Y.V. Romanov, Effekty Mnogomernosti v
Releynoy Zashchite Elektrichestvo 9, 48-54 (in Russian)
(2011)

5
E3S Web of Conferences 124, 01039 (2019) https://doi.org/10.1051/e3sconf/201912401039
SES-2019

[4] S.V. Ivanov, Y.Y. Lyamets, Metod

Informatsionnogo Analiza. Raspoznavaniye Zamykaniy
v Zadannoy Zone Dvukhtsepnoy Elektroperedachi
Izvestiya akademii nauk energetika 1, pp 47-57 (in
Russian) (2016)
[5] I.V. Nagay, V.I. Nagay, Postroyeniye
Mnogoparametricheskikh Rezervnykh Zashchit
Elektricheskikh Raspredelitel'nykh Setey 6–10 kV
Energetik 2, 18-21 (in Russian) (2013)
[6] M.V. Sharygin, A.L. Kulikov, The Effect of
Combining Information of Relay Protection Fault
Detectors Power Tech. and Eng. 52(5), 597-604 (2019)
[7] M.V. Sharygin, A.L. Kulikov, Automated
Calculation and Coordination of Protective Relay
Settings Power Tech. and Eng. 51(5), 593-601 (2018)
[8] A.B. Loskutov, M.V. Sharygin, A.L.
Kulikov, A.A. Loskutov, Ensuring Relay Selectivity in
Power Supply Systems using the Bayesian Method of
Hypothesis Testing IJAER 12(24), 15420-31 (2017)
[9] A.L. Kulikov, A.A. Loskutov, M. Mitrovic,
Method of automated synthesis of the logic part of relay
protection device which increases its sensitivity Int. Sc-
Tech Conf. ISEPC-2019 (23-24 May 2019, St.
Petersburg, Russia) (2019)
[10] J. McCalley, O. Oluwaseyi, V. Krishnan, R.
Dai, C. Singh, K. Jiang, System Protection Schemes:
Limitations, Risks, and Management PSERC 10-9 (2010)
[11] IEEE PSRC WG I-19 Redundancy
Considerations for Protective Relaying Systems (2008)
[12] M. Lukowicz, E. Rosolowski, Fault type
classification in high voltage power systems using
artificial neural networks Wroclaw University of
Technology (Poland) 1-6 (1996)
[13] M.M. Saha, E. Rosolowski, J. Izykowski,
2001 Artificial Intelligent Application to Power System
Protection ResearchGate 595-600
[14] Y.Y. Hong, P.C. ChangChian, Detection and
correction of distorted current transformer current using
wavelet transform and artificial intelligence Inst. Eng.
Technol. Gen. Transm. Distrib, 2(4), 566-75 (2008)
[15] N. Singh, H.K. Verma, S. Deb, Development
of ANN Based Relay Algorithm for Protection of Series
Compensated Transmission Line IOSR J of El. and
Elect. Eng. (IOSR-JEEE) 11(4), 74-82 (2016)
[16] S.R. Samantaray, I. Kamwa, J. Geza,
Decision tree based fault detection and classification in
distance relaying Int. J. of Engin. Intel. Sys. for Elect.
Eng. and Comm 18(2), 1-9 (2011)
[17] J. Azarakhsh, The Power Transformer
Differential Protection Using Decision Tree Bulletin de
la Société Royale des Sciences de Liège 86, 726-38
(2017)
[18] S. Netsanet, J. Zhang, D. Zheng, Bagged
Decision Tree Based Scheme Of Microgrid Protection
Using Windowed Fast Fourier and Wavelet Transforms
Electronics 7(61), 1-25 (2018)
[19] Ž. Zrno, Neke primjene normalne (Gaussove)
distribucije Zbornik radova Međimurskog veleučilišta u
Čakovcu, 2(2), 129-43 (in Serbian) (2011)
[20] H. Hemark, New Electrical Devices in
Existing Nuclear Power Plants REPORT 445,18 (2017)